Keep in mind, there is a reason that UDP was chosen for the standard. It is a lot lighter/skinnier and in probably 95%+ of the cases it works perfectly fine. You don't necessarily want to fatten up all of the traffic for everyone to fight an occasional problem. The problems tend to come in from misconfigurations of network devices or because people get overzealous tend to think of UDP as unimportant traffic (if its important, it will use TCP!!) so aren't afraid to configure devices to quickly drop it. That is old thinking completely due to old crappy networks. I fully expect to see more and more use of UDP especially in internal networks as the networking tech get more and more stable. The loss of the overhead required for maintaining the sessions is a good thing. It means less processing power for processing the extra info, less memory resources for maintaining all of the TCP connection state info, and less network traffic due to less data to go over the wire.
 
Think of it this way, right now AD works by sending a basic simple string stream across for most LDAP ops. Consider that someone decides that it would be much better if all info was XML encapsulated because it is was less likely to be confused or something like that and you get to a near doubling in the bytes on the wire and anything asking for the info has to now deal with all of the XML. I point this specific item out because there is a case where this occurs with AD Data that is returned over LDAP for replication metadata attributes. It is much smaller and tighter coming over as a binary pack, though definitely not as easy to use for some people.
 
I would be highly surprised if Microsoft changed this across the board for all kerberos packets instead of possibly just lowering the packetsize "again"[1] when it happens (but not to 1). What would be really cool is if they tried to work out the logic for it to figure out that is happening when a client is trying to logon and automatically kick the value down a little at a time until it works. That could probably all be done in far less time than it takes for people to spend 30 minutes logging on and then admins spending hours or days working out why.
 
I have hit this twice in the real world. The first time was with some very overzealous firmware settings on Cisco load balancers (CSMs) because "UDP traffic isn't important" which Cisco corrected with a patch and once when we had a DC that had somehow gotten the wrong network card device driver installed on it and it was dorking up all sorts of things.  
 
 
joe
 
 
[1] It went from 2000 to under 1500 in 2K to K3.
 
 
--
O'Reilly Active Directory Third Edition - http://www.joeware.net/win/ad3e.htm 
 
 
 


From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Myrick, Todd (NIH/CC/DNA) [E]
Sent: Wednesday, March 15, 2006 3:32 PM
To: ActiveDir@mail.activedir.org
Subject: RE: [ActiveDir] Communication across a trust...with firewalls

Something I wish they did when the product was released J

 

Todd

 


From: Olivarez, Sergio J Mr CTNOSC/GD-NS [mailto:[EMAIL PROTECTED]
Sent: Tuesday, March 14, 2006 5:14 PM
To: ActiveDir@mail.activedir.org
Subject: RE: [ActiveDir] Communication across a trust...with firewalls

 

Rocky, the reg hack that is referenced in Jeff’s response is one that I have seen Microsoft highly encourage and push towards customers.  I’ve spoken to a couple of Microsoft reps that say that this hack will be implemented in a future service pack\ patch because of all the problems it alleviates.  I really see no risk in making this change, all it does is force Kerberos over TCP.

 

Thanks... ... ... ...

Sergio J. Olivarez - Contractor

GD-NS


From: Jeff Salisbury [mailto:[EMAIL PROTECTED]
Sent: Tuesday, March 14, 2006 2:23 PM
To: ActiveDir@mail.activedir.org
Subject: RE: [ActiveDir] Communication across a trust...with firewalls

 

Rocky - This article explains why the fragmented UDP packets cause problems: http://support.microsoft.com/?id=244474 and how to modify the registry to force TCP. We run into this periodically, especially with users running a VPN tunnel across their home wireless network.

 

Jeff 

 


From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Rocky Habeeb
Sent: Tuesday, March 14, 2006 12:18 PM
To: ActiveDir@mail.activedir.org
Subject: RE: [ActiveDir] Communication across a trust...with firewalls

So ....  why does fragmentation cause a problem?  Packets are fragmented all the time in network traffic but stuff still works.  Are you saying credentialling packets can't be fragmented?

 

RH

___________________________________

 

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Myrick, Todd (NIH/CC/DNA) [E]
Sent: Tuesday, March 14, 2006 2:55 PM
To: ActiveDir@mail.activedir.org
Subject: RE: [ActiveDir] Communication across a trust...with firewalls

You might also want to investigate if you are using TCP or UDP packets with your authentication request.  By default Kerberos uses UDP, so a lot of firewalls will fragment the packets and cause authentication issues.

 

Todd Myrick

 


From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Almeida Pinto, Jorge de
Sent: Tuesday, March 14, 2006 2:47 PM
To: ActiveDir@mail.activedir.org
Subject: RE: [ActiveDir] Communication across a trust...with firewalls

 

lets say the structure is:

 

CLIENT-DOMAIN_A     .....     DC-DOMAIN_A    ......    DC-DOMAIN_B       ......   MEMBERSRV-DOMAIN_B

 

if NTLM is used the order of authentication is:

(1) CLIENT-DOMAIN_A wants to access MEMBERSRV-DOMAIN_B

(2) CLIENT-DOMAIN_A connects to MEMBERSRV-DOMAIN_B

(3) MEMBERSRV-DOMAIN_B connects to DC-DOMAIN_B and asks do you know: CLIENT-DOMAIN_A

(4) DC-DOMAIN_B says NO, but I do trust DOMAIN_A. Let me check.

(5) DC-DOMAIN_B connects to DC-DOMAIN_A and asks do you know: CLIENT-DOMAIN_A

(6) DC-DOMAIN_A says: yes, it's OK

(7) DC-DOMAIN_B sets up an access token for domain B for CLIENT-DOMAIN_A.

(8) CLIENT-DOMAIN_A accesses MEMBERSRV-DOMAIN_B

 

if KERBEROS is used the order of authentication is:

(1) CLIENT-DOMAIN_A wants to access MEMBERSRV-DOMAIN_B

(2) CLIENT-DOMAIN_A connects to DC-DOMAIN_A and asks for a ticket to access MEMBERSRV-DOMAIN_B

(3) DC-DOMAIN_A says: let me check, just a sec.

(4) DC-DOMAIN_A says: that server does not exist within the domain or the forest. However I do have a trust with DOMAIN_B. Go to DC-DOMAIN_B

(5) CLIENT-DOMAIN_A connects to DC-DOMAIN_B and asks for a ticket to access MEMBERSRV-DOMAIN_B

(6) DC-DOMAIN_B says: let me check, just a sec.

(7) DC-DOMAIN_B says: here's your ticket and access token. have fun

(8) CLIENT-DOMAIN_A accesses MEMBERSRV-DOMAIN_B

 

the problem is that only DC-DOMAIN_A and DC-DOMAIN_B can communicate through the firewall with each other. Other communication paths are not available or possible because of the firewall configuration.

 

Or did I miss something?

 

Met vriendelijke groeten / Kind regards,

Ing. Jorge de Almeida Pinto

Senior Infrastructure Consultant

MVP Windows Server - Directory Services

 

LogicaCMG Nederland B.V. (BU RTINC Eindhoven)

(         Tel     : +31-(0)40-29.57.777

(    Mobile     : +31-(0)6-26.26.62.80

*   E-mail      : [EMAIL PROTECTED]

 


From: [EMAIL PROTECTED] on behalf of [EMAIL PROTECTED]
Sent: Tue 2006-03-14 16:35
To: ActiveDir@mail.activedir.org
Subject: [ActiveDir] Communication across a trust...with firewalls

Within a domain, when a user’s credentials are presented to a member server, that member server communicates with the domain controller to validate the creds.

 

We have a cross-forest (cross–company; a divestiture) trust set up that we are testing.  A member server in the other forest/domain and across the firewall is having trouble authenticating credentials from our domain.  Their DC works fine.  Ports on the firewall are only opened for the two domain controllers (one on each side).

 

Here’s the question:  in order to validate the “foreign” credentials, should the member server be looking first to its own DC, or is it trying to cross the firewall to find our DC?  Based in the preliminary traffic sampling so far, I think that’s what is happening.  Is that normal/expected behavior?

 

TIA,

AL

Al Maurer
Service Manager, Naming and Authentication Services
IT | Information Technology
Agilent Technologies
(719) 590-2639; Telnet 590-2639
http://activedirectory.it.agilent.com

 

Confidential
This e-mail and any files transmitted with it are the property
of Belkin Corporation and/or its affiliates, are confidential,
and are intended solely for the use of the individual or
entity to whom this e-mail is addressed.  If you are not one
of the named recipients or otherwise have reason to believe
that you have received this e-mail in error, please notify the
sender and delete this message immediately from your computer.
Any other use, retention, dissemination, forwarding, printing
or copying of this e-mail is strictly prohibited.

Reply via email to